Chromatin of endoreduplicated pavement cells has greater range of movement than that of diploid guard cells in Arabidopsis thaliana

Fluorescence in situ hybridization analysis has shown that chromatin in eukaryote interphase cells including plants are fairly well organized(Cremer and Cremer, 2001; Fransz et al., 2002; Lysak et al., 2001). Studies using the lac-repressor (LacI)/green-fluorescent-protein (GFP) +lac-operator tagging system in budding yeast(Heun et al., 2001; Marshall et al., 1997a),dissected Drosophila tissues(Marshall et al., 1997a; Vazquez et al., 2001) and cultured mammalian cells (Chubb et al.,2002) showed that chromatin dynamics in interphase nuclei are generally constrained. This implies that chromosomes are anchored within the nucleus, at such sites as nuclear pores or nuclear lamina in fungi and animal cells (Chubb et al., 2002; Heun et al., 2001; Ishii et al., 2002). Chromatin dynamics in plant interphase cells has been poorly characterized in comparison.

We previously demonstrated the feasibility of detecting and tracking specific insertion sites in live Arabidopsis plants with the LacI-GFP+ lac-operator tagging system(Kato and Lam, 2001). In this approach, concatameric arrays of lac operators (256-mer, 10.1 kbp)were randomly inserted into chromosomes along with a glucocorticoid-inducible system that drives expression of a fusion protein that consists of GFP, LacI DNA-binding domain and the SV40 T-antigen nuclear localization signal(GFP-LacI/NLS; Fig. 1a). Upon induction of GFP-LacI/NLS protein, we can detect the tagged loci as GFP-binding fluorescent spots. We showed that addition of isopropyl-β-D-1-thiogalactopyranoside (IPTG), which inhibits LacI binding to the lac operator, significantly decreased the number of detectable GFP spots. We also showed that the induced GFP-LacI/NLS fusion proteins do not localize to specific sites when the lac operator array is absent from the genome. These controls demonstrated that there are few artefacts in our system, in which 1-2-week-old seedlings with a size range of 5-10 mm are directly mounted on microscope slides with water as the mounting medium(Fig. 2a). Under these conditions, Arabidopsis seedlings can remain viable for at least 24 hours and can grow normally upon their return to soil.

In this work, we measured chromatin dynamics in interphase plant cells. This is the first report of such studies for live plants, to our knowledge. Moreover, we tested whether chromatin dynamics change in different cell types. In Drosophila, the confinement area of spermatocyte nuclei(Vazquez et al., 2001) is much larger than that of stage 12 embryos(Marshall et al., 1997b). However, it is not known whether these differences are a reflection of cell type or the GFP-tagged loci examined. In this work, we compared chromatin dynamics of specific loci in different cell types of the same transgenic plants so that we can eliminate locus- and sample-related variations. We chose to compare guard cells and pavement cells for two reasons. First, the two cell types are both located on the epidermis of the seedlings and can be distinguished by their shapes (Fig. 2b). Therefore, we could observe the nuclei of these cells with minimal optical perturbation. Second, most, if not all, the guard cells are diploid whereas most pavement cells are polyploid owing to DNA endoreduplication (Melaragno et al.,1993). Therefore, we might be able to correlate the effects of endoreduplication with changes in chromatin dynamics.

Endoreduplication is a phenomenon in which chromosomes duplicate without any obvious condensation and decondensation steps, leading to the production of endopolyploid cells with enlarged nuclei. Endoreduplication is widely observed in the animal and plant kingdoms, although it is more common among insects and angiosperms that have small genomes(Nagl, 1976). Constant tissue-specific patterns of endoreduplication in different organs and cell types suggest that endoreduplication is an essential part of the developmental programs in post-mitotic cells (Mizukami,2001). In fact, ploidy levels affect gene expressions in budding yeast (Galitski et al., 1999)and plants (Lee and Chen,2001). It was of interest to us to study the effects of endoreduplication on chromatin dynamics because it could provide clues to the mechanisms that mediate changes in gene expression in polyploid cells(Edgar and Orr-Weaver, 2001; Madlung et al., 2002).

Second and fourth generation seedlings from the EL702C homozygous line(Kato and Lam, 2001) were used in this study. Seeds were germinated for 2-3 weeks in a plant growth chamber on 0.5× MS agar plates [2.1 g l^–1 Murashige and Skoog salts (Invitrogen, CA) and 0.7% bacto-agar, pH 6.0] containing 37 mg l^–1 of hygromycin prior to analysis.

Genomic DNA of homozygous EL702C Arabidopsis plant was extracted and digested with BglII and HindIII to release the T-DNA left border region from the lac repressor expression cassette together with genomic DNA flanking the insertion sites. DNA fragments of 2-5 kbp were used to construct the partial genomic DNA library with a ZAP Express, BamHI-predigested Gigapack cloning kit (Stratagene Cloning Systems,CA). The library (2×10⁵ pfu) was screened with radiolabeled T-DNA left border region DNA fragments(Kato and Lam, 2001).

BglII and HindIII-digested genomic DNA fragments of 2-5 kbp were eluted from an agarose gel and ligated to a BamHI cassette that is provided with the LA-PCR in vitro cloning kit (Takara, Japan). PCR was performed according to the manufacturer's protocol. Oligonucleotides(5′-ATGTCTCTGACCAGACACCCATCAACAGTA-3′ and 5′-GGGATCCTGTACTCCACCAAAGAAGAAGAG-3′) whose sequences are present in the T-DNA left border region were used as defined region-specific primers.

Surface-sterilized seeds were germinated on 0.5× MS agar plates. 1-2-week-old seedlings were transferred to fresh 0.5× MS agar plates and 40 μl of a 0.3 μM dexamethasone (Dex) solution was then dropped on each seedling. After 10-12 hours, the seedlings were placed between two coverslips under water. The coverslips were then placed on the microscope stage for microscopic imaging.

We followed the method described by Melaragno et al.(Melaragno et al., 1993). Homozygous EL702C seedlings were fixed in a solution of three parts 95%ethanol to one part glacial acetic acid for 2 hours at room temperature and then transferred to 70% ethanol. Fixed cells were then soaked in water and transferred into 0.5 M EDTA containing 5 μg ml^–1 of 4′,6′-diamidino-2-phenylindole (DAPI) for 1 hour. The stained seedlings were washed with water and set on the microscope slides as described above. Nuclear images were obtained with a 60× water-immersion objective lens and two-dimensional non-deconvolved nuclear images were used to obtain the fluorescence intensities. The mean intensity of three nuclei in root tip cells was used to compare the relative intensities of guard cell and pavement cell nuclei. Three independent samples were used.

A DeltaVision restoration microscope system (Applied Precision, WA)equipped with a TE200 microscope (Nikon, Japam) was used to observe nuclei in Dex-induced seedlings. 40 images at 0.2 μm z-axis steps were collected using a Nikon PlanApo 60× 1.2 NA water-immersion objective lens. The exposure times were 0.3-3.0 seconds and the filters used were:exciter, 436 nm/10 nm; emitter, 470 nm/30 nm; and a JP4 beamsplitter (Chroma,VT). The images presented were deconvolved based on a point-spread function data.

Volumes of the nuclei were estimated by measuring diffused GFP background fluorescence in the nuclei, which were used to measure chromatin movements, at the first time point. Deconvolved three-dimensional (3D) nuclear images were analysed to obtain the volumes. GFP signals were selected in each optical section and the volumes were calculated as a sum of GFP signal areas in a total of 40 sections.

Overall mean-square differences in distance between GFP spots<Δd²> plotted against elapsed time intervalsΔ t in guard-cell nuclei of cotyledons from EL702C homozygous plants are measured. When the data on the 12th (11 minutes) and 13th (12 minutes) time points were collected, the exciter shutter was closed from the 3rd to 10th points in order to minimize photobleaching of the samples.

Plants were fixed by treatment with ice-cold methanol for 5 minutes followed by ice-cold acetone for 30 seconds. The seedlings were immediately dried with paper towels and floated in a Petri dish with distilled water until it was set on the microscope slide.

The stacked images of nuclei were analysed by softWoRx software (Applied Precision, WA) on an Octane Workstation (Silicon Graphics, CA). The images were then processed by Adobe Photoshop 5.5 (Adobe Systems, CA) on a PowerMac G4 computer (Apple, CA) for the final images.

The number of integrated loci was characterized by Southern blot analysis,and the borders of the insertion sites were defined by sequencing respective integration sites using a genomic DNA library cloning strategy coupled with adaptor PCR. In the transgenic line designated as EL702C, we found three T-DNAs inserted into two linked loci on the upper arm of chromosome 3 with an interlocus distance of ∼4.2 Mbp (Fig. 1b). The insertion closer to the top end of the chromosome appears to be a simple insertion. The single T-DNA was integrated in a non-coding region ∼1.6 Mbp away from the top end of the chromosome. A gene that encodes an unclassified protein, the 26S proteasome AAA-ATPase subunit RPT5a[Arabidopsis Genome Initiative (AGI) annotation number: AT3g05520] is located 0.4 kbp from the left border (LB) of this T-DNA insert. The other neighboring gene, which encodes an unclassified protein [the α subunit of F-actin capping protein (AGI number: AT3g05530)], is located 2.0 kbp from the right border (RB) of this insert. The other two T-DNA inserts were inversely integrated in the same locus with a predicted 15 bp gap. An unidentified reading frame (URF) (AGI: AT3g16880) is disrupted by the inserts. One URF (AGI: AT3g16870) is located 1.9 kbp upstream of the LB for the insert,closer to the top of the chromosome arm, and another URF (AGI: AT3g16890) is located 0.7 kbp from the LB of the distal insert. Owing to the close proximity of these two inserts, we treat them as a single locus in our interpretation of microscopy observation in addition to the other single insertion, 4.2 Mbp away.

In Arabidopsis thaliana, ploidy levels for cells can vary from 2C(stomatal guard cells, sepal and petal epidermal cells) to 64C (leaf trichome)(Melaragno et al., 1993). Guard cells are components of stomata that regulate gas exchange and are distributed across the leaf and hypocotyl epidermis with characteristic spacing. These cells develop in pairs symmetrically from common mother cells and their cell cycle is arrested in G1 phase after differentiation(Melaragno et al., 1993). The shape of the differentiated cell is a uniformed crescent shape(Fig. 2b), whereas their nucleus is usually disk-like (Fig. 2d). In contrast to guard cells, pavement cells fill the surface of hypocotyl epidermis and grow in one direction without cell division,resulting in elongated, rectangular shapes(Fig. 2c). These cells contain a large nucleus that is usually elongated along the axis of the hypocotyl, and mature pavement cells have typically undergone several rounds of endoreduplication (Fig. 2e).

We first characterized the nuclei of guard cells and pavement cells in homozygous EL702C seedlings by fluorescence microscopy after Dex induction. In guard-cell nuclei, the spot numbers observed upon Dex treatment vary from none(not detected) to four (Fig. 2d, Fig. 3a);∼40% of the guard cells in homozygous lines showed two spots. In pavement cell nuclei, the spot numbers were more variable and ∼10% of cells showed up to eight spots (Fig. 2e, Fig. 3a).

To verify the expected ploidy difference between guard cells and pavement cells, the relative quantities of DNA in nuclei of these two cell types were measured by DNA-specific DAPI staining. Fluorescent intensities of each nucleus were compared with the average intensities of root-tip-cell nuclei, in which the ploidy level is known to be 2C(Melaragno et al., 1993). Fluorescent intensities of guard-cell nuclei in the homozygous EL702C line are the same as root-tip-cell nuclei (Fig. 3b). By contrast, DAPI fluorescence intensities in pavement cell nuclei are two to eight times higher than root-tip cells, and the average intensity is approximately five times higher than that of guard cells(Fig. 3b). These results are consistent with previous studies(Melaragno et al., 1993). The difference in nuclear volumes between guard cells and pavement cells was determined by calculating the volume of diffuse GFP background fluorescence caused by unbound fusion proteins in 3D models of the nuclei(Kato and Lam, 2001). In guard cells, nuclear volumes are ∼50 μm³, whereas the volumes of nuclei in pavement cells are ∼250 μm³ on average(Fig. 3c). The positive relationship between nuclear volume and ploidy levels is consistent with previous findings (Baluska,1990) and shows that the guard cells and the pavement cells we observed display the expected degree of ploidy. In this study, we did not measure the GFP spot numbers in the same samples that were used for DAPI staining because DAPI staining affected GFP-LacI/NLS induction and might also affect the viability of the samples. Therefore, we could not directly address the relationship between GFP spot numbers and ploidy level within the same nuclei.

To compare chromatin dynamics in diploid and endoreduplicated nuclei, we characterized the motion of the inserted loci in nuclei of guard cells and pavement cells in live EL702C Arabidopsis seedlings after Dex induction. Chromatin motions were represented by plotting the overall mean squared change in distance <Δd²> between two spots against elapsed time interval Δt(Marshall et al., 1997b). For two freely diffusing objects undergoing random-walk motion, a plot of<Δd²> against Δt should increase continuously if the movement is not constrained. If the movement is constrained to a certain area, <Δd²> should become independent of Δt over time and the plot will plateau.

First, we measured the chromatin movement in guard cells. We detected two spots in most of the analysed nuclei. We measured the distance between pairs of spots per nucleus as a function of time. As a total, we measured nine pairs of spots in a total of nine nuclei. The plot of measurements taken from live plants produced <Δd²> values that essentially remain unchanged in the range of 10 minutes(Fig. 4). We compared the plot to that observed with chemically fixed plants and found that the value of<Δd²> for live cells showed an approximately three times higher range at all time points measured, demonstrating that the movements are not due to experimental error or instrument noise. The asymptotic <Δd²> value of 0.03μm² (Fig. 4)indicates a mean change in distance between two spots of 0.17 μm. The maximum radius of the constrained area for guard cell nuclei, which was obtained by measuring Δd for various time points from independent experiments (data not shown) was 0.21 μm. The results indicate that the tagged loci in the transgenic line undergo diffusive movement in constrained areas within the nucleus. For particles undergoing 3D random walks with a diffusion constant D, it can be shown that a plot of<Δd²> against Δt should increase linearly with a slope of 4D(Marshall et al., 1997b). Thus, one can estimate the diffusion constant based on the plots. In Drosophila spermatocytes, diffusion coefficients of early G2 phase chromatin are 1.0×10^–3 μm²s^–1 and 0.94×10^–4 μm²s^–1 in late G2 phase. At the present time, with our experimental system, it takes ∼50 seconds to detect the tagged loci with high signal-to-noise ratios using 3D restoration microscopy. Thus, we are unable to perform more rapid time-lapse measurements of the relative distance between two tagged loci. We consequently could not measure precisely the earlier phase in the plot of <Δd²> versusΔ t in order accurately to determine the precise diffusion coefficient for individual tagged loci. However, because the asymptotic value of <Δd²> (0.03 μm²) was apparently reached in less than 1 minute, we estimate this would put a lower limit of 1.25×10^–4 μm²s^–1 for D, assuming that the plot of<Δd²> against Δt reaches a plateau at 1 minute in a linear fashion.

Next, we measured the chromatin movement of pavement cell nuclei by a plot of the overall mean squared change in distance<Δd²> between two spots for the elapsed time interval Δt (Fig. 4). The spot numbers varied between three and six spots in analysed nuclei. In most cases, `spot numbers – 1' pairs were measured for each of the nuclei studied. As a total, we measured 36 pairs of distances in ten nuclei. We also performed similar measurements with pavement cell nuclei that were chemically fixed to confirm that the different movements from guard cell nuclei are not caused by altered conditions when dealing with different cell types in different parts of the plant. Plots of measurements made with pavement cell nuclei show that <Δd²>increases until 7 minutes (Δt) and reached a plateau of 0.10μm², which indicates a mean change in distance between the spots of 0.32 μm. The maximum radius of the confinement area, which was also obtained by measuring Δd for various time points from independent experiments (data not shown), was 0.44 μm. Diffusion coefficients calculated from the slope is 5.35×10^–5μm² s^–1. We did not detect any correlation for distances between the pairs of spots and the changes (i.e. there is no evidence that spots close to each other move less than spots that are further apart) in this work.

The Student's t-test for <Δd²>values using a two-tailed distribution between the samples of diploid guard cells and endoreduplicated pavement cells at 9 minutes and 10 minutes gave P values of 0.0050 and 0.0054, respectively. This suggested that the values of <Δd²> at the plateau are significantly different. Our results indicate that chromatin movement of endoreduplicated pavement cell nuclei might have a lower diffusion coefficient(<2.3 times) but larger constrained areas (6.6 times larger on average and 9.2 times as a maximum) than guard cell nuclei.

The GFP spot numbers in the observed nuclei varied from measurement to measurement. This suggests that some of the tagged loci might associate stochastically with each other, in a manner similar to a previously reported phenomenon in budding yeast (Aragon-Alcaide and Strunnikov, 2000). We tested whether spot size or fluorescence intensity is correlated with the spot numbers but we could not find positive correlations in either guard cells and pavement cells (data not shown). This could be due to differences of spot positions in the z-axis. Meanwhile, we found that the rate of association in pavement cells is higher than that in guard cells. We do not yet know whether the higher rate of association in the pavement cells is caused by the stochastic association of the insertion loci or whether each replicated chromatid remains cohered in nuclei that undergo endoreduplication.

However, we can estimate the coherence probabilities of multiple chromatids in endoreduplicated cells by comparing the average GFP spot numbers and the average DAPI fluorescence intensities in endoreduplicated pavement cells versus diploid guard cells. If all chromatids are separated, the ratio of GFP spot numbers in pavement cells to guard cells would be the same as that for DAPI fluorescence intensities. By contrast, if all chromatids remain cohered,the GFP spot numbers would not increase (guard-cell to pavement-cell ratio of 1). Because the increment of GFP spot number ratio (2.5 times on average) is half of that for the increase in DAPI fluorescence intensities (five times on average) in this study (Fig. 3a,b), we predict that half of the chromatids might remain cohered while the other half are separated in the endoreduplicated pavement cells. This is consistent with the recent fluorescence in situ hybridization (FISH)observation using fixed 4C nuclei with single-locus probes for chromosome 4. The observation suggests that ∼50% of the chromatids in endoreduplicated nuclei are not cohered (M. Klatte, V. Schubert and I. Schubert, personal communication).

In both guard cells and pavement cells, chromatin shows constrained diffusion movement (Fig. 4). This suggests that plant chromatin is tethered by static interactions within nuclei, as found in other organisms (Chubb et al., 2002; Heun et al.,2001; Ishii et al.,2002). The result allows studies of chromatin dynamics, which have been done in yeast and animals, to be extended to plants with regards to the mechanism of interchromosomal interactions. Lamin proteins are thought to provide scaffolding structures that mediate the attachment of chromatin to the nuclear envelope in animal cells (Gant et al., 1999; Lopez-Soler et al.,2001). However, lamins have not been identified in plants or the budding yeast Saccharomyces cerevisiae(Meier, 2001). This fact and our present finding suggest that plants and yeast might have a distinct set of nuclear anchoring components that functionally replace lamins of animal cells.

One of the features of endoreduplicated nuclei is increase in nuclear volume (Baluska, 1990). Although the biological consequences of the enlarged nucleus on nuclear processes such as transcription are not clear(Edgar and Orr-Weaver, 2001), a simple explanation of this feature is that nuclear volume is proportional to ploidy levels to maintain a similar 3D space per haploid genome. The measured increase in nuclear volume in pavement cells is approximately the same as the estimated increase in DNA content relative to diploid guard cells(Fig. 3b,c). Moreover, half of the chromatids in endoreduplicated pavement cells are thought to be separated,as discussed above. These features suggest that pavement-cell nuclei indeed contain a similar (twofold difference in maximum) 3D space per haploid genome as guard-cell nuclei.

By contrast, we found that the apparent confinement areas for chromatin movement of endoreduplicated pavement-cell nuclei are over six times larger than those found in guard-cell nuclei. This finding suggests that the chromatin in endoreduplicated pavement cells has larger free spaces of diffusion than that in guard-cell nuclei even though the chromosome space per haploid genome is similar. What makes the confinement areas larger in pavement nuclei than in guard cells? One simple explanation could be a reduction in the concentration of proteins involved in chromatin organization in the larger pavement-cell nuclei. This might result in an enlarged free space for chromosome motion in these endoreduplicated cells, because of a decrease in the number of anchoring sites or interactions for chromatin within the nucleus. In fact, daughter nuclei of early G1 phase yeast cells, which are 40%smaller than the mother-cell nuclei, show five-times-less-frequent large rapid chromatin movements than mother cells in the same culture(Heun et al., 2001).

Another observation that lends support to this explanation is the fact that endoreduplicated cells of maize root show a decrease in staining with Fast Green FCF, which stains basic proteins within the nuclei(Balusuka and Kubica, 1992). An alternative explanation is that the compositions of the chromatin-organizing proteins are different between the two types of nuclei. Although the functional significance of matrix-attachment motifs in both cis- and trans-acting elements remains unclear, the importance of structural features of cis-acting elements such as a narrow minor groove in A/T-rich DNA has been suggested (Kas et al., 1989). It can be hypothesized that endoreduplicated nuclei have a different type or sets of chromatin-organizing proteins from diploid nuclei. Their different sequence affinity for anchoring chromatin and changes in their quantities or composition could result in alteration of chromatin dynamics. Therefore, it would be interesting to determine whether diploid and endoreduplicated nuclei have different biochemical compositions in matrix proteins.

Changes of gene expression associated with ploidy differences in budding yeast is not caused by alterations in growth rate or viability but instead by an increase of DNA content (Galitski et al., 1999). Allotetraploid (hybrid genome)-dependent changes in gene expression of Arabidopsis have also been reported(Lee and Chen, 2001). In the study, genes silenced in Arabidopsis suecica, an allotetraploid derived from Arabidopsis thaliana and Cardaminopsis arenosa,were investigated with an amplified fragment-length polymorphism (AFLP) cDNA display method. The result indicates epigenetic regulation of orthologous genes in polyploid genomes.

Epigenetic mechanisms can involve interchromosomal interactions. For example, paramutation can silence allelic genes in heterozygous conditions(Sidorenko and Peterson, 2001)whereas transvection can change gene expression by combining an enhancer on one chromosome and a promoter on another chromosome(Wu and Morris, 1999). Marshall explained the relationship between interchomosmal interactions and chromatin diffusion confinement as follows(Marshall, 2002). If two loci are to interact physically, they must be located within the same 3D space in the nucleus. However, because of non-random nuclear organization, not every pair of loci will be equally likely to interact. Owing to constrained diffusion, only those loci whose regions of confinement overlap are capable of interacting; pairs of loci whose confinement regions do not overlap are unable to interact because they cannot diffuse into contact.

Based on this functional model, we propose a model of developmental control of gene expression in endoreduplicated cells: although the volume of endoreduplicated nucleus is larger than that of diploid nucleus, the nuclear space per haploid genome or chromosome territories stays similar. Therefore,locus densities per nuclear space would be similar to those in diploid nuclei. Endoreduplicated nuclei allow additional or reduced intra- and interchromatin interactions by providing larger free spaces for chromatin diffusion. The expansion in the effective confinement area in the same 3D space per haploid genome (compared with the diploid nucleus) might arise from a decrease in the concentration or a change in the properties of nuclear structural proteins. In this manner, endoreduplicated cells might achieve a different gene expression pattern from diploid cells through an epigenetic mechanism. This model might also apply to the evolutionary strategy for species that carry small genomes to modulate gene expression patterns in specific cell types, which has been achieved in other species by varying DNA contents(Nagl, 1976).

This work is supported by a grant from the Plant Genome Research Program of NSF (#0077617) to E.L. We would like to thank W. Marshall for constructive suggestions on the analysis of chromosome movements and useful comments, M. Lawton for careful reading and correction of the manuscript, and N.-H. Chua for the pTA7002 vector used to construct the EL702 vector.